Eye tracking is more sensitive than skin conductance response in detecting mild environmental stimuli

Abstract

The skin conductance (SC) and eye tracking data are two potential arousal-related psychophysiological signals that can serve as the interoceptive unconditioned response to aversive stimuli (e.g. electric shocks). The current research investigates the sensitivity of these signals in detecting mild electric shock by decoding the hidden arousal and interoceptive awareness (IA) states. While well-established frameworks exist to decode the arousal state from the SC signal, there is a lack of a systematic approach that decodes the IA state from pupillometry and eye gaze measurements. We extract the physiological-based features from eye tracking data to recover the IA-related neural activity. Employing a Bayesian filtering framework, we decode the IA state in fear conditioning and extinction experiments where mild electric shock is used. We independently decode the underlying arousal state using binary and marked point process (MPP) observations derived from concurrently collected SC data. Eight of 11 subjects present a significantly (P-value $<0.001$ ) higher IA state in trials that were always accompanied by electric shock ( $\mathrm{CS}+\mathrm{US}+$ ) compared to trials that were never accompanied by electric shock ( $\mathrm{CS}-$ ). According to the decoded SC-based arousal state, only five (binary observation) and four (MPP observation) subjects present a significantly higher arousal state in $\mathrm{CS}+\mathrm{US}+$ trials than $\mathrm{CS}-$ trials. In conclusion, the decoded hidden brain state from eye tracking data better agrees with the presented mild stimuli. Tracking IA state from eye tracking data can lead to the development of contactless monitors for neuropsychiatric and neurodegenerative disorders.

Keywords: arousal; eye tracking; interoceptive awareness; pavlovian conditioning; skin conductance response.

Fig. 1.

An example of the recovered ANS activation from the SC signal (a measure of EDA) for one subject. The left column represents the complete recorded data, and the right column shows a random window of the experiment. The sub-panels, from top to bottom, present: A) the raw SC signal; B) the recovered ANS activation.

Fig. 2.

An example of the eye-based feature extraction steps and the formation of the IA-related observation for one subject. The left column represents the complete recorded data, and the right column shows a random window of the experiment. The sub-panels, from top to bottom, present: A) the onset of the fixation (blue impulses) and the fixation duration (the amplitude of each impulse); B) the average pupil size (derived from both eyes); C) the pupil size derivative (dilation and constriction speed); D) the average velocity (derived from both eyes); E) the average acceleration (derived from both eyes); F) the recovered IA-related events from the eye (blue impulses).

Fig. 3.

The event-related potential-like (ERP-like) analysis in the fear conditioning and extinction experiments. The sub-panels of the figure present: A) the epoch of the average velocity and its 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject; B) the epoch of the pupil size and its 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject; C) the epoch of SC signal, and its 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject; D) the epoch of the decoded IA state from IA-related neural activities derived from the eye tracking data (eye-based), and its 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject; E) the epoch of the decoded arousal state from ANS activations and their amplitudes (MPP) derived from the EDA measurements, and the 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject. F) the epoch of the decoded arousal state from arousal events (binary), and its 95$\mathrm{\%}$ confidence limits across $\mathrm{CS}+\mathrm{US}+$ trials (red), $\mathrm{CS}+\mathrm{US}-$ trials (yellow), and $\mathrm{CS}-$ trials (blue) for an example subject. G) the box plots for the epoch of the decoded IA state from the eye tracking data (eye-based) across $\mathrm{CS}+\mathrm{US}+$ trials (red box), $\mathrm{CS}+\mathrm{US}-$ trials (yellow box), and $\mathrm{CS}-$ trials (blue box) for all subjects; H) the box plots for the epoch of the decoded arousal state (MPP-based) from the EDA measurements across $\mathrm{CS}+\mathrm{US}+$ trials (red box), $\mathrm{CS}+\mathrm{US}-$ trials (yellow box), and $\mathrm{CS}-$ trials (blue box) for all subjects; I) the box plots for the epoch of the decoded arousal state (binary-based) from the EDA measurements across $\mathrm{CS}+\mathrm{US}+$ trials (red box), $\mathrm{CS}+\mathrm{US}-$ trials (yellow box), and $\mathrm{CS}-$ trials (blue box) for all subjects. The *** is used to indicate $P<0.001$ where the findings are statistically significant.

Fig. 4.

An example of the decoded IA and arousal states for one subject. The left column represents the estimates of the eye-based IA using a binary observation (IA-related neural activity), the middle column corresponds to the estimates of the EDA-based arousal derived from MPP observation, and the right column depicts the EDA-based estimates of arousal from a binary observation. The sub-panels, from top to bottom, depict: A) the physiological signal of interest (e.g. pupil size or SC); B) the applied observation vector; C) the decoded state trajectory its 95$\mathrm{\%}$ confidence limits; D) the probability of event occurrence; E) the internal brain state index (HII or HAI).